Insurance claims adjusters just earned an unwanted title: the most anti-AI workers in America. Glassdoor combed through employee reviews and found that 98 percent of adjuster comments mentioning AI were negative between June 2025 and May 2026. Across the wider insurance industry, the negativity rate was 81 percent, still the third-worst of any sector Glassdoor measured.
This is not a small group of complainers. It is a profession that has watched its own job market shrink while being told the shrinkage is a feature, not a bug. Employment in claims adjusting fell about 21 percent year over year through May 2026, compared with a much smaller 2.5 percent drop across insurance carriers overall. Entry-level postings for the role have fallen close to 50 percent since early 2024.
The adjusters interviewed for this story describe a pattern that should worry any manager thinking about automating customer-facing work. AI systems misclassify claims, invent details in medical summaries, and generate estimates that later turn out wrong. When that happens, the human adjuster still has to catch the mistake, explain it to an angry customer, and redo the work. The promised time savings turn into extra labor, plus the reputational cost of looking incompetent for an error a machine made.
Not every company is failing at this. Lemonade built its business around an AI-first claims process from the start rather than bolting AI onto an existing human workflow, and its chatbot now handles the bulk of initial claim filings with a large share of claims requiring no human step at all. The difference between Lemonade and the frustrated adjusters in legacy insurers is not the technology. It is whether the process was redesigned around AI or whether AI was dropped into a system built for humans and then left for the humans to fix when it breaks.
There is a legal dimension too, and it matters beyond insurance. Courts have already started allowing discovery into whether insurers used AI to deny claims without proper human review, treating an uncaught AI error as potential evidence of bad faith. Any company using AI to make decisions that affect customers, denials, approvals, pricing, is building similar exposure if a human is not meaningfully checking the output.
The quieter problem is what happens to the career ladder. Entry-level claims work has always been how adjusters learn the judgment they need for complex cases later. Cut that training ground now, and insurers may find themselves short of experienced staff in five or ten years, right when catastrophe claims and disputed cases still need a human who knows what they are doing.
Investors are not waiting to find out. AI insurance startups have raised tens of millions of dollars in the past year alone, on the assumption that today's rough automation gets smoother fast. Adjusters, for their part, doubt it will happen fast enough to save their jobs, or their sanity, in the meantime.